420 SECTIONAL TRANSACTIONS.— J 



Dr. G. Seth. — The development of infant behaviour : a study of the 

 responses of infants from 20 fo 52 weeks in age (12.15). 



Afternoon. 



Dr. P. B. Ballard. — Intelligence tests and secondary school entrance 

 examinations (2.0). 



Mr. T. E. Stubbins. — Prognostic value of school examinations (2.45). 



In an attempt to determine which part of an Entrance Scholarship 

 Examination most successfully predicted success in a School Certificate 

 Examination, coefficients of correlation between each part of the former and 

 nine subjects taken by the same (180) pupils in the latter were obtained. 

 Owing to the existence of intercorrelations between the parts of the Scholar- 

 ship Examination, these coefficients lead to conclusions of doubtful value. 

 Regression Equations, in which allowance is made for such intercorrelations, 

 were therefore worked out. 



The equations are of the form : 



Xi = aXi + bxg + CA'4 + dx^ + k 

 where Xi represents a candidate's score in a School Certificate subject 



X2, X3, Xi, and x^ the same candidate's marks in the various Scholar- 

 ship subjects, 

 ^ is a constant, 

 and a, b, c and d are the regression coefficients, being the weights to be 

 assigned to the appropriate scholarship mark. 



A high positive weight indicates that the corresponding Scholarship 

 mark has high predictive value in the School Certificate subject denoted 



A coefficient of multiple correlation gives the relationship existing between 

 the marks gained in the School Certificate Examination and those which 

 would be obtained by substituting his Scholarship marks in the equation. 



The entrance examination was of interest, in that it comprised attainment 

 tests in English and Arithmetic, an Intelligence Test, and estimates sub- 

 mitted by the Headmasters of the Elementary Schools attended by the 

 pupils. 



Dr. D. A. Walker. — Answer-patterns (3.30). 



It is a fact well known to examiners that the nature of a test paper deter- 

 mines to a certain extent the way in which scores in the test will be distri- 

 buted. The nature and extent of this predetermination of score-scatter 

 by the type of paper set has been the subject of the investigation reported 

 in this paper. So far the research has been confined to tests where each item 

 carries unit score. It is then possible to construct for each test an answer- 

 pattern, which is a table of the frequencies of correct answers to the various 

 items, these being placed in order of difficulty. In the particular case 

 where each candidate's score is made up of answers to the easiest items, it 

 can be shown that there is an exact relationship between the answer-pattern 

 and the score-scatter of the test. In the more general case there still persists 

 some measure of relationship, and this has been studied by experimental 

 methods. A particular case of this relationship is the production by a diffi- 

 cult test paper of a positively skewed score-scatter, spacing out the best 

 candidates. 



